Forum Discussion
Abdulmohsen_Mut
Nimbostratus
NimbostratusDeny Large file download using HTTP::conent-Length
Hi,
I am trying to implement something similar to the one in the thread titled :redirect POSTs bigger than 1MB
http://devcentral.f5.com/Default.aspx?tabid=28&forumid=5&postid=5411&view=topic
We need to intercept responses to large file download request and close the HTTP session before the download starts. However, i am worried that this will impact the performance of our proxies as thousands of users will be going through the BIG-IP in front of the proxy servers.
Also, I need to send the user a small warning html that tell him that downloading files > 50MB is not allowed, by re-writing the payload.
I am thinking of something like:
when HTTP_RESPONSE {
set clen [HTTP::header Content-Length]
if { $clen > 50,000,000} {
HTTP::payload replace 0 [HTTP::payload length] $warning_response
HTTP::close
}
log local0. "HTTP_RESPONSE CLEN= $clen "
}
Please advise on how to achieve this with minimum impact on performance. I am worried that enabling this iRule will cause the BIG-IP to read every HTTP meassage to look for Content-Length. Is there a way to read only required messages ??
Your support is highly appreciated.
3 Replies
- It looks like you're on the right track so far. Have you tested this rule yet? Is it behaving the way you'd expect?
As far as only opening certain packets goes, that's really not feasible. You're deciding which packets to disallow/alter based on information inside the packet, so you have to open all of them to see if they match or not.
The up side is, the ones that aren't over your maximum allowed conent length will skip most of the rule, and should cause very little impact to your system.
-Colin 
Abdulmohsen_MutThanks Colin.
Would you please provide a typical complete iRule that i can use safely for testing. I don't know how would i write the payload with a complete html page. Do i provide a link or do i have to copy the entire html in my code.
Your support is highly appreciated.
uniAltocumulus
HTTP::respond might be a better choice than HTTP::payload replace. You get to give them an appropriate response code and I expect is more efficient.
Also, I think you should remove the commas from 50,000,000 as this would make it a textual comparison rather than numeric.
